io.github.avaazquezz/mcp-qdrant

qdrant_points_upsert

DestructiveIdempotent

Insert or replace points in a Qdrant collection, with optional insert-only, update-only, or filtered update modes to control overwrites.

Instructions

Insert or replace points (id + vector + payload) in a collection.

    `vector` matches the collection's vector config: a plain list for a
    single unnamed vector; a list of lists for an unnamed multivector;
    or a dict keyed by vector name for named vectors, where each value
    is a dense list, a multivector, or a sparse vector
    (`{"indices": [...], "values": [...]}`).

    `update_mode` makes the write conditional: `insert_only` never
    overwrites an existing id, `update_only` never creates a new one
    (default `upsert` does both). `update_filter` only overwrites existing
    points whose current payload matches the filter. `wait=false`
    returns before the write is applied.

    Fails with a clear error if the collection doesn't exist.

    Example: {"collection_name": "docs", "points": [
        {"id": 1, "vector": [0.1, 0.2, 0.3, 0.4], "payload": {"city": "ny"}}
    ]}
    Example (named dense + sparse): {"collection_name": "docs", "points": [
        {"id": 1, "vector": {"dense": [0.1, 0.2, 0.3, 0.4],
            "sparse": {"indices": [1, 7], "values": [0.5, 0.2]}}}
    ]}

Input Schema

Name	Required	Description	Default
`wait`	No
`points`	Yes
`update_mode`	No
`update_filter`	No
`collection_name`	Yes

Output Schema

Name	Required	Description	Default
`status`	Yes
`operation_id`	No	Sequential number of the operation

Schema Changelog

Changes observed during successful MCP inspections.

Changed9 schema fields changedv2.6.0Oct 25, 2026

addedInput schema / $defs / PointInput / additionalProperties
```
Added value: +false
```

changedInput schema / $defs / PointInput / description

Previous value: -"A point to upsert: a plain id + vector + payload — never `Document`,\n`Image`, or `InferenceObject` (qdrant-client's server-side embedding\ninference variants), which would contradict this server's principle of\nnever generating embeddings itself."New value: +"A point to upsert: id + vector(s) + payload. See `PointVectorValue`\nfor the accepted vector shapes."

addedInput schema / $defs / PointInput / properties / vector / anyOf

Added value: +[
+  {
+    "items": {
+      "type": "number"
+    },
+    "type": "array"
+  },
+  {
+    "items": {
+      "items": {
+        "type": "number"
+      },
+      "type": "array"
+    },
+    "type": "array"
+  },
+  {
+    "additionalProperties": {
+      "anyOf": [
+        {
+          "items": {
+            "type": "number"
+          },
+          "type": "array"
+        },
+        {
+          "$ref": "#/$defs/SparseVector"
+        },
+        {
+          "items": {
+            "items": {
+              "type": "number"
+            },
+            "type": "array"
+          },
+          "type": "array"
+        }
+      ]
+    },
+    "type": "object"
+  }
+]

removedInput schema / $defs / PointInput / properties / vector / items
```
Removed value: -{
-  "type": "number"
-}
```
removedInput schema / $defs / PointInput / properties / vector / type
```
Removed value: -"array"
```

addedInput schema / $defs / SparseVector

Added value: +{
+  "additionalProperties": false,
+  "description": "Sparse vector structure",
+  "properties": {
+    "indices": {
+      "description": "Indices must be unique",
+      "items": {
+        "type": "integer"
+      },
+      "title": "Indices",
+      "type": "array"
+    },
+    "values": {
+      "description": "Values and indices must be the same length",
+      "items": {
+        "type": "number"
+      },
+      "title": "Values",
+      "type": "array"
+    }
+  },
+  "required": [
+    "indices",
+    "values"
+  ],
+  "title": "SparseVector",
+  "type": "object"
+}

addedInput schema / properties / update_filter

Added value: +{
+  "anyOf": [
+    {
+      "additionalProperties": false,
+      "description": "Qdrant payload filter. Combine conditions with `must` (all hold), `should` (any holds), `must_not` (none holds) or `min_should` ({conditions, min_count}); each takes a list of conditions or a single one. A condition is one of: {\"key\": k, \"match\": M} where M is exactly one of {\"value\": v}, {\"any\": [..]}, {\"except\": [..]}, {\"text\": s}, {\"text_any\": s}, {\"phrase\": s}, {\"prefix\": s}; {\"key\": k, \"range\": {\"lt\"|\"lte\"|\"gt\"|\"gte\": number or datetime}}; {\"key\": k, \"values_count\": {\"lt\"|\"lte\"|\"gt\"|\"gte\": n}}; {\"key\": k, \"geo_radius\": {\"center\": {\"lat\", \"lon\"}, \"radius\": meters}}; {\"key\": k, \"geo_bounding_box\": {\"top_left\", \"bottom_right\"}}; {\"key\": k, \"geo_polygon\": {\"exterior\", \"interiors\"}}; {\"is_empty\": {\"key\": k}}; {\"is_null\": {\"key\": k}}; {\"has_id\": [ids]}; {\"has_vector\": name}; {\"slice\": {\"total\": n, \"index\": i}}; {\"nested\": {\"key\": k, \"filter\": <filter>}}; or a nested filter. Keys for nested objects use dots (`a.b`); objects inside arrays use `a[].b`.",
+      "properties": {
+        "min_should": {
+          "anyOf": [
+            {
+              "properties": {
+                "conditions": {
+                  "items": {
+                    "type": "object"
+                  },
+                  "type": "array"
+                },
+                "min_count": {
+                  "type": "integer"
+                }
+              },
+              "required": [
+                "conditions",
+                "min_count"
+              ],
+              "type": "object"
+            },
+            {
+              "type": "null"
+            }
+          ]
+        },
+        "must": {
+          "anyOf": [
+            {
+              "anyOf": [
+                {
+                  "items": {
+                    "type": "object"
+                  },
+                  "type": "array"
+                },
+                {
+                  "type": "object"
+                }
+              ]
+            },
+            {
+              "type": "null"
+            }
+          ]
+        },
+        "must_not": {
+          "anyOf": [
+            {
+              "anyOf": [
+                {
+                  "items": {
+                    "type": "object"
+                  },
+                  "type": "array"
+                },
+                {
+                  "type": "object"
+                }
+              ]
+            },
+            {
+              "type": "null"
+            }
+          ]
+        },
+        "should": {
+          "anyOf": [
+            {
+              "anyOf": [
+                {
+                  "items": {
+                    "type": "object"
+                  },
+                  "type": "array"
+                },
+                {
+                  "type": "object"
+                }
+              ]
+            },
+            {
+              "type": "null"
+            }
+          ]
+        }
+      },
+      "type": "object"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null
+}

addedInput schema / properties / update_mode

Added value: +{
+  "anyOf": [
+    {
+      "enum": [
+        "upsert",
+        "insert_only",
+        "update_only"
+      ],
+      "type": "string"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null,
+  "title": "Update Mode"
+}

addedInput schema / properties / wait

Added value: +{
+  "default": true,
+  "title": "Wait",
+  "type": "boolean"
+}

First observedv1.1.1Sep 1, 2026

TDQS

A4.5/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare destructiveHint=true, idempotentHint=true and readOnlyHint=false, so the safety profile is known; the description adds genuine value on top by explaining the conditional write modes, the deferred-write behaviour of `wait=false`, and the failure mode when the collection is missing. What it does not resolve is the key destructive question for a replace operation: whether an existing point's payload is merged or wholly overwritten when the same id is upserted.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with the one-line purpose, then parameter semantics, then examples — a sensible ordering where every block earns its place. It is on the long side, and the two full JSON examples repeat structure that the prose above already conveyed, but the second example is what makes the named dense+sparse shape concrete.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a five-parameter mutation tool with an output schema and rich annotations, the description covers everything an agent needs to call it correctly: vector shape selection, mode semantics, the deferral caveat, and the missing-collection error. Return values are handled by the output schema, so no gap remains.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Top-level schema description coverage is effectively 0% for `vector`, so the description carries the burden and does so thoroughly — plain list for a single unnamed vector, list of lists for a multivector, dict keyed by vector name for named vectors with sparse `{indices, values}` support. It also defines each `update_mode` value and the meaning of `update_filter` and `wait=false`, with two worked examples covering the dense and the dense+sparse cases.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb pair and resource: 'Insert or replace points (id + vector + payload) in a collection.' The verb-plus-resource combination ('upsert points') is unambiguous against siblings like qdrant_points_get, qdrant_points_delete and qdrant_points_scroll, which perform entirely different operations on the same resource.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives real conditional guidance on the modes: `insert_only` never overwrites an existing id, `update_only` never creates a new one, `update_filter` only overwrites matching points, and `wait=false` returns before the write is applied. It does not, however, tell the agent when to prefer a sibling tool (e.g. qdrant_points_delete for removals or qdrant_query for reads), so routing between tools is left to inference.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server